Federated Clustering Defense: Mitigating Data Exposure in Collaborative Learning

The development of intelligent industrial applications is being driven globally by advancements in the next generation of Internet of Things (IoT) technologies. At the same time, artificial intelligence (AI) technologies like machine learning and deep learning are becoming more widely used. Conventional machine learning models primarily rely on large data sets. However, the vast data sets generated by network-edge devices are expensive, inefficient to acquire and handle, and they pose major privacy hazards. Federated learning (FL) is a novel paradigm for statistical model training on distributed edge networks that allows data to take part in the training of federated models without being localized. Traditional machine learning issues with low data usage, data privacy, and information security brought on by data isolation can be resolved with this method. We propose a novel Federated Clustering with Differential Privacy (FedCDP) algorithm designed to efficiently model user behavior in an unsupervised manner. This algorithm focuses on optimizing computation and storage requirements while minimizing data leakage through robust differential privacy mechanisms.